Professional Documents
Culture Documents
Fenrg 10 1005749
Fenrg 10 1005749
Fenrg 10 1005749
REVIEWED BY
algorithm
Bicheng Yan,
King Abdullah University of Science and
Technology, Saudi Arabia
Congcong Li 1, Chaoqiang Fang 2, Yougen Huang 3, Hailong Zuo 3,
Hari Ganesh, Zhang Zhang 1 and Shuoliang Wang 1*
Indian Institute of Technology
Gandhinagar, India 1
School of Energy, China University of Geosciences, Beijing, China, 2CNPC Logging Company Limited
Hui Zhao, Geological Research Institute, Beijing, China, 3CNPC Exploration and Development Research Institute
Yangtze University, China of Changqing Oilfield Branch, Xi’an, China
*CORRESPONDENCE
Shuoliang Wang,
wangshuoliang@cugb.edu.cn
SPECIALTY SECTION A significant amount of bypassed oil resources often remain in a mature
This article was submitted to Process
waterflooding reservoir because of non-uniform sweep caused by natural
and Energy Systems Engineering,
a section of the journal complexities of a subsurface reservoir and improper management of the
Frontiers in Energy Research reservoir. Infill drilling is one of the most attractive options for increasing oil
RECEIVED 28 July 2022 recovery in consequence of its operational simplicity, low risk and promising
ACCEPTED 27 October 2022
results. Determining optimal infill well placements in heterogeneous mature
PUBLISHED 16 November 2022
reservoirs is a critical and challenging task that has a significant impact on the
CITATION
Li C, Fang C, Huang Y, Zuo H, Zhang Z
recovery performance and economic revenue of subsurface remaining oil resources.
and Wang S (2022), Infill well placement An integrated framework is constructed to attain best-obtained optimal location and
optimization for secondary completion of infill wells in multi-layer mature oil reservoirs. The placement of an infill
development of waterflooding oilfields
with SPSA algorithm. vertical well is parameterized in terms of two sets of variables that define the location
Front. Energy Res. 10:1005749. and completion respectively. A variant of SPSA algorithm is used to solve the defined
doi: 10.3389/fenrg.2022.1005749
optimization problem. The performance of the proposed algorithm is first tested for
COPYRIGHT the joint optimization of well location and completion of an injection well using a
© 2022 Li, Fang, Huang, Zuo, Zhang and
Wang. This is an open-access article synthetic model. The results show that the algorithm with average SPSA gradients
distributed under the terms of the outperforms the single SPSA gradient method both in solution and convergence rate.
Creative Commons Attribution License
Besides, there are two plateaus on the performance curve of all algorithms: on the
(CC BY). The use, distribution or
reproduction in other forums is first plateau, each algorithm is approaching to its optimal well location with relatively
permitted, provided the original little change on the completion parameters, while on the second plateau, each
author(s) and the copyright owner(s) are
credited and that the original algorithm obtains the corresponding optimal completions. A complex
publication in this journal is cited, in heterogeneous reservoir model is then constructed by using the data of a mature
accordance with accepted academic
oil reservoir in Shengli Oilfield in China to design an optimal 10 years’ infill drilling
practice. No use, distribution or
reproduction is permitted which does program. Four vertical production wells are placed in the oil-rich regions and both
not comply with these terms. simultaneous and sequential algorithms are tried to obtain their optimal locations and
completions. The performances of simultaneous joint optimization and sequential
joint optimization are compared and as a result it is recommended to use sequential
joint optimization as the optimization algorithm in the integrated framework.
KEYWORDS
1 Introduction located in, the reservoir layer to be penetrated and the length of
the perforated or open section of a well will together influence the
About 92% of oil reservoirs in China are characterized by development performance greatly. Hence, the main objective of
continental clastic sediments with complex vertical and lateral infill well placement optimization should be optimizing not only
heterogeneities and high levels of crude oil viscosity (Han, 2010). the areal drainage pattern but also the vertical scheme.
Waterflooding with uniformly spaced well pattern has always Optimizing both well location and perforation along
been the main development approach for Chinese oilfields. After formation intervals could help maximize the control reserve of
decades of producing, most of the major oilfields have entered the a single well, homogenize the inflow-velocity profile and prevent
late stage of waterflooding life cycle featured by high water-cut early water breakthrough and rapid water-cut increase in high-
(greater than 80%) and high recovery of recoverable reserves permeability completion intervals, which are all critically
(greater than 70%). Although most of mature oilfields are important for enhancing oil recovery and improving
encountering such terrible situations, they still contribute to a economic revenue of mature oilfields. However, there could be
great proportion of crude oil supply in China. According to a large number of possible candidate locations for a new well. To
official statistics, more than 70% of the total national oil search through and evaluate all the possible locations is not
production are from mature oilfields at present (Han, 2010). practically feasible, particularly for high resolution geologic
Production practice shows that there still is a great potential to models consisting of multimillion cells. For large scale field
improve the recovery of high water-cut oilfields. The decreasing applications, a practical method is needed to mitigate the
trend in discovering new conventional hydrocarbon resources is computational burden associated with the possibly large
necessitating optimal recovery from existing reserves in a more number of simulation runs.
cost-effective manner. Traditionally, optimization of well placement has been done
Due to strong reservoir heterogeneities, the remaining oil by using quality maps that indicate which regions of the reservoir
distribution generally presents a feature of “highly scattered on have not been adequately swept by the injected water (Badru,
the whole and relatively abundant in local areas” when the 2003; Kharghoria et al., 2003; da Cruz et al., 2004; Guimaraes
comprehensive water-cut in oilfields is beyond 80% (Han, et al., 2005; Taware et al., 2012). The quality maps are typically
2007). The two major methods to improve water sweep based on static properties such as permeability, porosity,
efficiency that were once applied effectively at lower water-cut structure and net thickness, and dynamic properties such as
stage, including subdivision of development intervals and remaining oil, pressure, well productivity and cumulative oil
uniform infilling wells, began to show little or even no effects. production. Although this method is convenient for application,
This is mainly because the achievements of the above methods at it cannot properly place wells in locations that take advantage of
lower water-cut stage of development are owed to the large-area long-term high oil saturation or respond to the dynamics of
distributions of remaining oil in relatively low-to-moderate reservoir fluid flow over a long period of time. Simply placing
permeability layers at that time, but the current condition wells based on saturation or quality maps may not guarantee long
becomes very different in that the overall remaining oil is term profitability of the project. In recent years, various
highly scattered even in low-to-moderate permeability layers. simulation-based optimization techniques have also been
In this case, uniform infilling wells are not as efficient as before. introduced to automate the process of attaining the optimal
The initial water-cut in most newly drilled infilling wells can be as well locations. There are many challenges in the field of
high as 80% and few even higher than 90% (Han, 2010). In order automatic well placement optimization, such as variable
to further improve the recovery of remaining oil reserves, it needs design, objective function, constraints formulation,
to drill individualized infill wells rather than uniform infill wells. optimization algorithm, geological uncertainty, computational
Determining appropriate well placements of producers and cost and so on. We will just discuss a few of them for brevity in
injectors is an important aspect of overall development strategy this paper.
of any field. Well placement optimization becomes particularly One important aspect of automatic well placement
noteworthy in mature oilfields where an enormous amount of approaches is the description of well location and/or well
pore volumes are occupied by water and both reservoir geological trajectory. As for well placement problem, the actual physical
characteristic and remaining oil distribution are extremely problem of determining well location and trajectory is a
complex. Thus, new wells have to be drilled based on a continuous problem, while the completions of a well in the
comprehensive understanding of reservoir description. In formation are commonly introduced into the numerical
Chinese mature oilfields, each wellbore may penetrate many simulation model as integral indices of the connected
prospective reservoir layers that can range, depending on the gridblocks. Thus, to parameterize the location and, in general
field size and number of reservoirs, from a few into the hundreds. the trajectory of the wells, both discrete variables (Yeten et al.,
However, technical and reservoir management considerations 2003; Bangerth et al., 2006; Ozdogan and Horne, 2006; Emerick
usually make every wellbore capable of producing from only a et al., 2009; Bellout et al., 2012; Li et al., 2013) and continuous
limited completion interval. The region that an infill well to be real-valued variables (Onwunalu and Durlofsky, 2010;
Bouzarkouna et al., 2012; Forouzanfar et al., 2012; Forouzanfar optimization problems where the gradient information is not
and Reynolds, 2013; Nwankwor et al., 2013; Forouzanfar et al., available (Spall, 1992; Spall, 1994; Spall, 1998; Spall, 2000).
2016) are used. For a discrete optimization problem, decision Various kinds of SPSA algorithms have been successfully
variables are well connection gridblock indices. The center point applied in several engineering fields. Early applications of
of the well is moved from one gridblock to another at each SPSA algorithm in petroleum engineering field were
iteration of whatever optimization algorithm is used. For a performed by Bangerth et al. (Bangerth et al., 2006) and
continuous optimization problem, the well trajectory is Gao et al. (Gao et al., 2007). In Bangerth et al. (Bangerth
generally parameterized in a more real but complex way, et al., 2006), a variant of SPSA (integer SPSA) is introduced
which could make the method uneasy to implement. to solve the well placement problem. The authors compared
Another important aspect of automatic well placement and analyzed the efficiency, effectiveness, and reliability of
approaches is the optimization algorithm. To solve the several optimization algorithms, including the simultaneous
optimal well placement problem, most researchers have perturbation stochastic approximation (SPSA), finite
utilized heuristic global optimization methods such as genetic difference gradient (FDG), very fast simulated annealing
algorithm (GA) (Montes et al., 2001; Yeten et al., 2003; Ozdogan (VFSA) and genetic algorithm (GA), for the well
and Horne, 2006; Emerick et al., 2009; Salmachi et al., 2013), placement problem by a set of numerical waterflooding
differential evolution (DE) (Afshari et al., 2015; Awotunde, experiments and found that none of these algorithms
2016), simulated annealing (Sa) (Beckner and Song, 1995), guarantees to find the optimal solution, but both SPSA
particle swarm optimization (PSO) (Onwunalu and Durlofsky, and VFSA are very efficient in finding nearly optimal well
2010; Bouzarkouna et al., 2013; Afshari et al., 2014; Forouzanfar locations with a high probability. In Gao et al. (Gao et al.,
et al., 2016), the covariance matrix adaptation-evolution strategy 2007), both SPSA and adaptive SPSA are applied to the
(CMA-ES) algorithm (Bouzarkouna et al., 2012; Jesmani et al., history matching problem. The authors presented
2016a; Forouzanfar et al., 2016) and some hybrid algorithms modifications for improvement in the convergence
(Nwankwor et al., 2013). Although these methods have the ability behavior of the SPSA algorithm for history matching and
to avoid local solutions, their convergence to the global solution compared its performance to the steepest descent, gradual
is heuristic in natural and typically require a very large number of deformation and LBFGS algorithm and concluded that
reservoir simulation runs in an optimization loop. Moreover, to although the convergence properties of the SPSA
account for model uncertainty, it is often need to evaluate the algorithm are not nearly as good as the LBFGS methods,
performance of well placements over multiple geological the SPSA algorithm is not simulator specific and it can be
realizations, which makes them too computationally expensive coupled easily with any commercial reservoir simulator to do
and thus, may not be well-suited for large scale field applications. automatic history matching. Since the two primary works,
Some local search optimization algorithms were also applied to the SPSA algorithm has been widely used to solve various
the well placement optimization problem such as gradient-based optimization problems in petroleum engineering field
method algorithms with adjoint gradient (Wang et al., 2007; (Wang et al., 2009; Li and Reynolds, 2011; Do et al., 2012;
Sarma and Chen, 2008; Zandvliet et al., 2008; Vlemmix et al., Do and Reynolds, 2013; Li et al., 2013; Jesmani et al., 2016b;
2009; Li and Jafarpour, 2012; Forouzanfar and Reynolds, 2014), Pouladi et al., 2020).
methods on the basis of the simultaneous perturbation stochastic For simulation-based well placement optimization
approximation (SPSA) (Bangerth et al., 2006; Li et al., 2013; approaches, the reliability of the solution depends
Jesmani et al., 2016b), derivative-free methods (Forouzanfar seriously on the main parameters defining the geological
et al., 2012; Forouzanfar and Reynolds, 2013; Isebor et al., model, including formation structure, porosity and
2014a; Isebor et al., 2014b) and generalized pattern search permeability distributions, and fluid contacts. Due to
(GPS) (Bellout et al., 2012). Among these local search limited reservoir description knowledge, these parameters
algorithms, gradient-based method has the superior are all uncertain and generally described by some probability
computational efficiency. However, one must calculate the density functions. Management of the effect of geological
gradient of the objective function and the constraints with uncertainty on well placement optimization is another
respect to the optimization variables to apply the efficient research area that is gaining more and more consideration
gradient-based method. The adjoint for computing these from researchers (Ozdogan and Horne, 2006; Li and
gradients is not commonly provided in commercial reservoir Reynolds, 2011; Li et al., 2013; Chang et al., 2015; Jesmani
simulators, which makes the adjoint methods are difficult to et al., 2016b; Temizel et al., 2018). A common practical
implement and require access to the simulator source code. approach for taking into account the uncertainty in well
An alternative approach to gradient-based method is the placement problem is to approximate the probability
stochastic optimization method. In particular, the simultaneous distributions with multiple equally probable geological
perturbation stochastic approximation (SPSA) has been realizations, and then optimize the expected value of the
developed for solving large dimensional multivariate objective function over an ensemble of the model
assigned to discrete grid blocks in the model, namely “connecting 2.1.2 Joint well placement optimization problem
gridblocks (I, J, K)”. This implies that for a given number of infill Based on the parameterization method for well placement,
wells to be optimized, Ninf ill , the dimension of the set of well the problem considered here can be given as the following
location variables, nw , is determined, that is, nw 2Ninf ill . formulation:
However, the dimension of the set of well completion
^ ^c arg
w, min f(s(m, w, c, t), m, w, c) (3)
variables, 2nw,c , is related not only to the number of infill w∈ZNw ,c∈RNc
wells but also to the perforation intervals being optimized of each g(s(m, w, c, t), m, w, c) 0 (4)
well. This makes the optimization problem more complicated to wl ≤ w ≤ wu (5)
define and to solve, because the number, the position and the cl ≤ c ≤ cu (6)
length of the well completion in multi layers are all the
optimization variables. In order to simply the optimization where f(·) is a specified objective function; s is the state of the
problem, we use the “well productivity index (WPI) reservoir at time t; w denotes the discrete well location variables
multiplier” instead of “connecting gridblocks” in reservoir with w u represents its upper bound and wl the lower bound; c is
simulators (Eclipse E100) as the optimization variables, the continuous completion variables with cu represents its upper
because the inflow performance between the wellbore and the bound and cl the lower bound; g(s(m, w, c, t), m, w, c) denotes
connecting gridblock is calculated by WPI and WPI multiplier the discretized multiphase flow equation with m represents the
can take into consideration the partially-penetrated length of a reservoir model input parameters (e.g. petrophysical and fluid
connecting gridblock. We define WPI multiplier as η Δl/Δz in properties).
this paper, where Δl is the connecting length of a single gridblock
with the well and Δz is the thickness of that gridblock. It is worth
to mention that, to our knowledge, the concept of WPI multiplier 2.2 Objective function
has been applied by Forouzanfar et al. (Forouzanfar et al., 2012)
for a different purpose. In Forouzanfar et al. (Forouzanfar et al., In the hydrocarbon production problems, the objective
2012), the placement of a vertical and horizontal well was function can be any user-specified performance measure, such
described as four continuous parameters, (xw , yw , zw , lw ) and as cumulative oil production or recovery factor and net present
WPI multiplier was used to modify well productivity indices in value (NPV). This paper considers the negative of the reservoir
the reservoir simulator to account for the location of the NPV increment due to infilling wells as the minimization
centerline of a well that does not have to be at the center of a objective function f(·). In general, the NPV is described as
gridblock when the optimal well placement is formulated as a the summation of present values in a time interval when present
continuous optimization problem. However, the authors did not values are calculated based on incoming and outflows cash,
consider multi-segment perforated intervals which are which is calculated by below equation:
commonly demanded for the redevelopment of mature oilfield Nt
CFn
in China. NPV tn /365
(7)
The infill well is initially assumed to connect with each layer n1 (1 + d)
Nt Nprod Ninj
Δtn xk+1 xk − ak g^k (xk ) (9)
NPV ⎡⎢⎣ ro qno,i − rwp qnwp,i − rwi qnwi,j ⎤⎥⎦ tn /365
n1 i1 j1 (1 + d) where g^k (xk ) is a stochastic approximation of ∇f(xk ) at the kth
− Niw ciw iteration and ak is a nonnegative scalar gain coefficient. The
(8) iterative form in Eq. 9 represents the steepest descent algorithm if
g^k (xk ) is replaced by ∇f(xk ).
where, ro , rwp and rwi are the oil price, cost of water production The central finite difference method is usually used to
disposal and cost of water injection per unit volume respectively; approximate the gradient by perturbing the elements of the
qno,i and qnwp,i are the average oil and water production rates of the decision variable vector one at a time and evaluating the
ith producer over the nth time interval; qnwi,j is the average water objective function for each perturbation, which can be written as
injection rate of the jth injector over the nth time interval; ciw is
y(xk + ck ei ) − y(xk − ck ei )
the cost of drilling per well; Nprod and Ninj , respectively, denote g^ki (xk ) (10)
the total number of both existing and infill production wells and 2ck
injection wells; Niw is the number of infill wells. where ei denotes a unit vector with its ith entry equal to unity and
In this formulation, the NPV increment objective function all other entries equal to zero; ck is a small positive scalar referred
depends on the oil production rate qo , the water production rate to as the perturbation size.
qwp , the water injection rate qwi and the number of infill wells. It is easy to see that 2p function evaluations are required to
The reservoir simulation model can be used to forecast the approximate the gradient using the finite-difference method,
production performance for an infill project with fixed which makes the method increasingly inefficient for
number and placement of infill wells, and then the revenue of application as the problem dimension p grows. The SPSA
the project can be estimated using the economic objective algorithm provides an efficient alternative by perturbing all
function. Increment of net present value compared with the vector elements simultaneously to obtain a stochastic gradient
value evaluated without infill wells means drilling additional approximation g^k (xk ) that is defined as
wells adds value to the current production from mature oil
y(xk + ck Δk ) − y(xk − ck Δk ) −1 T
reservoir and project should be economically acceptable. g^k (xk ) × Δ−1 −1
k1 Δk2 . . . Δkp
2ck
(11)
Although the gradient approximation in Eq. 11 requires less parameter scaling and constraint handling process, a simple
function evaluations, it is important to consider the number of line search procedure and a blocking step.
iterations required for effective convergence of the algorithm. Firstly, because the well placement and completion
The SPSA algorithm converges to a local optimum point under optimization formulation is a constrained optimization
some conditions on the SPSA parameters (gain sequences), problem and the elements of optimization variable set x has
smoothness of the objective function close to the optimum, and very different magnitudes, the variables are scaled using a linear
when the properties regarding the perturbation distribution Δk transformation with their upper and lower bounds. For the ith
discussed above are fulfilled (Jesmani et al., 2016b). The variable xi , we define the transformed new variable ui such that
conditions for the gain sequences are as follows: xi − xm,i
ui f ori 1, 2, L, Nx (15)
∞ ∞
xr,i
ak > 0, ck > 0, ak → 0, ck → 0, ak ∞, a2k c2k < ∞ (12)
k0 k1 where xm,i xu,i2+xl,i and xr,i xu,i2−xl,i with xl,i and xu,i denoting
the lower and upper bounds for the ith control variables. When
To satisfy the above conditions, Spall (Gao et al., 2007) xi → xu,i , ui → 1 and when xi → xl,i , ui → − 1. With this
provided some useful guidelines to select the perturbation size transformation, the optimization algorithm may simply works
ck and gain step length ak for SPSA, which are defined within [−1, 1]. However, since SPSA is an unconstrained
respectively as follows: optimization algorithm, it needs to incorporate a constraint
a handling procedure. In this work, a bounding operator and a
ak (13)
(A + k + 1)β rounding operator suggested by Bangerth et al. (Bangerth et al.,
c 2006) are applied to handle inequality constraints and the integer
ck (14)
(k + 1)γ problem of well location variables, respectively.
In addition, as the gradients are stochastic and approximated,
where a, A, c, β, and γ are positive real numbers which should
the descent direction might not have sufficient accuracy and the
satisfy the conditions of A ≥ 0, β − 2γ > 0 and 3γ − (β/2) > 0. The
magnitudes of the gradient may change significantly for each new
choice of these parameters can have some influence on the
generation. To avoid this problem and guarantee an appropriate
performance of the SPSA algorithm. In Spall (Gao et al.,
step size that minimizes the objective function, we also
2007), the commonly recommended values for β and γ are
implement the averaged stochastic gradient as a search
0.602 and 0.101, respectively. Moreover, it is suggested to set
direction, i.e.
A at or around 10% of the maximum number of expected or
allowed iterations k max , that is, A 0.1k max . After specifying A 1 g
N